davecb at Canada.Sun.COM
Wed Feb 3 21:05:32 GMT 1999
> What NetBench actually measures in a large part is the
> speed of the client memory cache :-) :-) :-).
> The only reasonable benchmark would be one based on
> open code that generated the CIFS calls via direct
> socket calls - not going through a Windows redirector
> (which is broken in *so many* subtle and not-so-subtle
> ways :-). The smbtorture code would be a good start
Literally what I was thinking of, for my
next book project.
I'm rereading the code and papers from the nfsstone->laddis
historical period, collecting half-baked ideas, and
studying up on how to cook them properly.
The more-baked-than-not parts:
1. Based on smbtorture-style calls
2. Driven from one of two distributions
of user activities: one pure nfsstone (ie,
actually a mix dating back to musbus), and
another based on the mix in netbench (as
reverse analyzed by netapp in the "
Adding Response Time Measurement of CIFS
File Server Performance to NetBench" paper.
3. In boringly simple c.
The flaky ideas are things like "yeilds comparable
numbers to XXX", which requires both some tuning knobs
and a real understanding of the XXX mixes in question.
David Collier-Brown, | Always do right. This will gratify some people
185 Ellerslie Ave., | and astonish the rest. -- Mark Twain
Willowdale, Ontario | http://java.science.yorku.ca/~davecb
Work: (905) 477-0437 Home: (416) 223-8968 Email: davecb at canada.sun.com
More information about the samba-technical